Abstract
The paper presents a new clustering scheme for wireless sensor networks based on semantic properties. Unlike current clustering schemes where signal strength and neighbouring arc, the most used criteria in the cluster formation, in this work, semantic properties are also taken in consideration in the process of cluster formation. Nodes inside a same cluster arc, organized like a hyper tree where the cluster-head is the root. This hyper tree organization allows a layered data aggregation, avoids the cluster-head overload and energy dissipation, and makes the network more robust against links and nodes failures. We evaluate this work by simulations and show that our approach is more efficient and more energy conserative than other clustering schemes.